Chart Selection and Statistical Graphics

Introduction

Data visualization is not primarily the production of attractive graphics; it is the design of visual arguments that allow people to detect patterns, compare alternatives, assess uncertainty, and make decisions. In information-intensive organizations, an inappropriate chart can conceal an important trend, exaggerate a difference, or encourage a conclusion unsupported by the data. Chart selection is therefore a problem of analytical reasoning as much as graphic design.

This chapter develops a systematic approach to selecting and designing tables and statistical graphics. It focuses on five recurring analytical purposes: comparison, distribution, relationship, composition, and ranking. It also examines visual encoding, scale design, statistical graphics, uncertainty, overplotting, accessibility, ethics, and the use of software and AI-assisted tools. The central principle is that the form of a display should follow the structure of the question and the measurement properties of the data.

Learning Objectives

After studying this chapter, students should be able to:

1. Match analytical questions and data types to suitable tables and charts.

2. Explain how visual encodings such as position, length, angle, area, colour, and shape affect interpretation.

3. Select appropriate displays for comparison, distribution, relationship, composition, and ranking.

4. Design statistically honest graphics with suitable scales, labels, annotations, and uncertainty indicators.

5. Diagnose clutter, distortion, overplotting, and misleading conventions.

6. Evaluate charts critically in technical, managerial, ethical, and governance contexts.

7. Use a structured workflow to design and validate a visualization for a real decision.

1. The Analytical Purpose Comes First

A visualization should answer a question. The same dataset may legitimately require several different displays because different displays support different forms of reasoning. For example, monthly sales data may be shown as a line chart to reveal temporal movement, a dot plot to compare branches, and a box plot to examine variation across regions.

Five broad purposes are especially useful:

• Comparison: Which category, group, or time period is larger or smaller?
• Distribution: How are values spread, concentrated, skewed, or outlying?
• Relationship: How does one variable vary with another?
• Composition: How does a total consist of parts, and how do those parts change?
• Ranking: What is the ordered position of each item?

These purposes overlap, but the distinction helps prevent habitual chart selection. A pie chart, for example, may communicate a simple composition at one point in time, but it is usually weak for comparing many categories or comparing composition across multiple periods.

1.1 Data structure and measurement scale

Chart choice depends on the variables being displayed. A nominal variable identifies categories without an inherent order, such as department or product type. An ordinal variable has an order but not necessarily equal intervals, such as satisfaction levels from poor to excellent. An interval variable has meaningful differences but no meaningful absolute zero, such as temperature in degrees Celsius. A ratio variable has a meaningful zero, such as revenue, distance, or elapsed time.

Variables may also be classified as:

• Quantitative: numerical measurements on which arithmetic operations may be meaningful.
• Categorical: labels or groups.
• Discrete: countable values, such as number of incidents.
• Continuous: measurements that can theoretically take values over an interval.
• Temporal: observations indexed by time.
• Spatial: observations associated with locations or regions.

The data structure includes more than variable type. The designer must ask whether observations are independent, repeated for the same entities, nested within groups, sampled from a population, or generated as a time series. A line chart connecting unrelated categories can imply continuity that does not exist; a map can imply geographic meaning when the data are merely assigned to administrative areas.

2. Visual Encoding and Perceptual Effectiveness

A chart translates data values into visual properties called encodings. Common encodings include position on a common scale, length, angle, area, colour intensity, colour hue, shape, and line connection. These encodings are not equally precise.

For quantitative comparison, position along a shared scale is generally among the most accurate encodings. Length is also effective, especially when bars share a baseline. Angle is less precise, which is one reason pie charts become difficult to read with many segments. Area and volume are still harder to compare because viewers do not estimate them linearly. Colour hue is useful for distinguishing categories, while colour intensity can represent ordered magnitude if the scale is perceptually appropriate.

A useful hierarchy is:

1. Position on a common scale.

2. Position on non-aligned scales.

3. Length.

4. Direction and angle.

5. Area.

6. Volume and three-dimensional form.

7. Colour and texture for quantitative precision.

This hierarchy is not an absolute law. Context, familiarity, accessibility, and task affect performance. Nevertheless, it explains why a sorted dot plot may communicate differences more accurately than a decorated radial chart.

2.1 Marks, channels, and the data-ink principle

A mark is a visible graphical object, such as a point, line, bar, or area. A channel is a visual property used to encode a variable. A chart should allocate prominent channels to the most important analytical variables and use secondary channels sparingly.

The data-ink principle recommends maximizing the proportion of visual elements that communicate data rather than decoration. This does not mean removing every non-data element. Gridlines, reference bands, annotations, and explanatory labels can improve interpretation. The relevant question is whether an element assists reading or merely competes for attention.

2.2 Pre-attentive features and attention

Some visual properties are detected rapidly before conscious sequential inspection. Position, orientation, size, and colour contrast can direct attention. This is valuable for highlighting an exception or target value, but it creates ethical responsibility: a designer can make a minor difference appear important simply by using a bright colour or oversized marker.

Emphasis should therefore correspond to analytical importance. If one point is highlighted, the chart should explain why—for example, because it represents the current period, a regulatory threshold, or a statistically unusual observation.

3. Tables: Precision Before Pattern

Tables are often the best choice when exact values, multiple dimensions, or auditability matter. They support lookup and detailed comparison better than most charts. A table should have a clear title, meaningful row and column labels, consistent units, appropriate rounding, and an indication of missing or suppressed values.

3.1 Frequency and contingency tables

A frequency table counts observations in categories or intervals. A contingency table cross-classifies two categorical variables. For example, an organization may cross-tabulate service rating by customer segment. Row percentages answer a different question from column percentages, so the basis of each percentage must be stated explicitly.

A table should not present excessive decimal precision. If measurements are reliable only to the nearest unit, displaying six decimal places falsely suggests accuracy. Totals should reconcile, and percentages should be checked for rounding discrepancies.

3.2 When a table is preferable

Use a table when:

• users need exact values;
• there are many categories or variables;
• small differences are important;
• the audience must verify or reproduce calculations;
• values include confidence intervals, denominators, or notes that would clutter a chart.

A hybrid display can combine a chart with a compact table, but duplication should serve a purpose. A dashboard that repeats every value in both a chart and a table may increase visual burden without improving decisions.

4. Charts for Comparison and Ranking

4.1 Bar charts

Bar charts are appropriate for comparing quantities across discrete categories. Bars should normally begin at zero when their lengths encode absolute magnitude. Categories should usually be sorted by value for ranking, unless a meaningful order exists, such as chronology, process sequence, or geographic arrangement.

Horizontal bars are preferable when category labels are long or when there are many categories. A consistent scale and common baseline are essential. Avoid decorative pictograms whose areas or volumes do not correspond transparently to values.

A truncated baseline can be acceptable in some cases, particularly for dot plots or carefully labelled point estimates, but it is risky for bars because viewers interpret bar length from the baseline. If a nonzero baseline is necessary, it should be clearly marked and the design should not imply more difference than exists.

4.2 Dot plots and interval displays

Dot plots place values on a shared quantitative scale and are often more efficient than bars. They are especially useful for comparing estimates, showing rankings, and displaying confidence intervals. A Cleveland dot plot, for example, can show multiple measures for categories without the visual mass of bars.

For paired comparisons, a slope chart or connected dot plot may show change between two conditions. Such charts should be used only when the pairing is meaningful and the number of categories is limited. Too many crossing lines create a “spaghetti” display.

4.3 Ranking displays

Ranking is a special form of comparison in which order is central. A sorted horizontal bar chart, dot plot, or lollipop chart can communicate rank effectively. However, rank alone may conceal magnitude. Two items may be adjacent in rank but nearly identical, while a large gap may separate other ranks. Therefore, labels, reference lines, or confidence intervals may be needed.

Rankings based on composite scores require particular caution. The chart should identify the dimensions, weights, normalization method, time period, and missing-data treatment used to construct the score. A precise-looking rank can be unstable when differences are small or uncertainty is large.

5. Charts for Distribution

A distribution describes how observations are allocated across values. It includes central tendency, spread, skewness, multimodality, gaps, and outliers. A single average rarely communicates the distribution adequately.

5.1 Histograms

A histogram divides a quantitative variable into adjacent intervals called bins and displays the count or density in each bin. Bin width influences the visible pattern: very wide bins can hide modes, while very narrow bins can create visual noise. The choice should be justified by the analytical purpose and, where appropriate, tested through sensitivity analysis.

If all bins have equal width, bar height may represent count or frequency density depending on the design. When bin widths differ, the area of each rectangle—not merely its height—must represent the count. This is a frequent source of statistical misinterpretation.

5.2 Box plots

A box plot summarizes a distribution using quartiles, a median, and a rule for displaying potential outliers. The interquartile range is

IQR = Q3 − Q1,

where Q1 is the first quartile and Q3 is the third quartile. A common convention marks observations below Q1 − 1.5(IQR) or above Q3 + 1.5(IQR) as potential outliers. This is a display convention, not a universal test for abnormality or data error. The numerical result depends on the quartile definition and sample context.

Box plots are compact and useful for comparing distributions across groups, but they can conceal multimodality, sample size, and the number of observations. Overlaying jittered points, a violin density, or a sample-size label can provide additional information, provided the display remains readable.

5.3 Density plots, empirical distributions, and ridgelines

A density plot estimates the shape of a distribution by smoothing observations. The smoothing parameter affects apparent peaks and should not be treated as an objective fact. Density plots are useful for comparing groups with sufficient data, but they may mislead with small samples or very unequal group sizes.

An empirical cumulative distribution function (ECDF) plots the proportion of observations less than or equal to each value. It avoids binning and can support direct probability comparisons. ECDFs are particularly useful when tails, thresholds, or stochastic dominance are important.

6. Charts for Relationships

6.1 Scatter plots

A scatter plot displays paired quantitative observations using position on two axes. It can reveal direction, form, strength, clusters, and outliers. A fitted line may summarize a model, but it should not be mistaken for proof of causation.

The correlation coefficient measures linear association under its stated assumptions; it does not capture every nonlinear relationship and can be strongly affected by outliers or restricted ranges. A scatter plot should therefore precede or accompany correlation claims.

Overplotting occurs when many points overlap. Remedies include transparency, smaller marks, jitter for discrete values, hexagonal binning, contour displays, sampling with disclosure, or aggregation. Each remedy changes what is visible. Transparency can reveal dense regions but may produce dark patches that are difficult to quantify.

6.2 Time-series graphics

Line charts are appropriate when observations have a meaningful ordered sequence, especially time. The horizontal axis should preserve temporal order and indicate the sampling interval. Connecting observations implies a trajectory between them; this is reasonable for many measurements but not for unrelated categories.

Time-series charts require attention to seasonality, missing periods, changes in definition, and irregular sampling. A moving average can reveal broad movement but may hide short-term variation and introduce edge effects. It should be labelled as a derived series rather than presented as raw data.

Dual axes are particularly hazardous. They can make unrelated variables appear to move together because each axis is independently scaled. If two measures must be compared, indexed values, small multiples, or standardized units are usually more transparent.

6.3 Relationships among many variables

A scatterplot matrix can show pairwise relationships, while a heat map can summarize a matrix of values such as correlations or errors. These displays require careful ordering, colour scales, and annotation. Correlation matrices should identify the correlation measure and sample basis; a strong association may be driven by confounding, common time trends, or a small number of observations.

7. Charts for Composition

Composition concerns how a whole is divided into parts. The first design question is whether the parts are mutually exclusive and collectively exhaustive. If categories overlap, a conventional part-to-whole chart is inappropriate unless the overlap is explicitly represented.

7.1 Stacked bars and area charts

Stacked bars work well for comparing totals and, with a common baseline, the first segment across categories. Interior segments are harder to compare because their baselines vary. A 100-percent stacked bar chart compares proportions but removes information about absolute totals. It should therefore not be used when both size and share are decision-relevant unless totals are shown separately.

Stacked area charts can communicate changing composition over an ordered sequence, but the movement of interior layers is difficult to interpret. They are most suitable when the overall pattern matters more than precise category comparison.

7.2 Pie and donut charts

A pie chart can be acceptable for a small number of parts that sum to a meaningful whole and have clearly different shares. It becomes weak when segments are numerous, similar in size, or compared across multiple pies. Sorted bars usually support more accurate comparisons. Donut charts remove the central reference point and do not solve the perceptual limitations of angle and area.

7.3 Waterfalls, treemaps, and mosaics

Waterfall charts show sequential contributions to a total, such as revenue less costs leading to profit. They require a clear additive logic and careful labelling. Treemaps encode values as nested rectangles and can display hierarchical composition, but area comparison is difficult and small items may be unreadable. Mosaic plots represent contingency-table structure through area and are useful for categorical association, but they require explanation for non-specialist audiences.

8. Statistical Honesty: Scale, Uncertainty, and Context

A truthful chart can still mislead if it omits denominators, uncertainty, or relevant context. Rates should be distinguished from counts. Comparing the number of incidents across regions without considering population size can produce a false ranking. Whenever possible, state the denominator, exposure period, and population at risk.

8.1 Axes and transformations

Linear scales preserve additive differences. Logarithmic scales represent multiplicative differences and are useful when values span several orders of magnitude or when proportional change matters. A logarithmic axis must be labelled clearly; zero and negative values cannot be represented directly on an ordinary logarithmic scale.

Percent change and percentage-point change are not interchangeable. If a proportion rises from 20% to 30%, the increase is 10 percentage points and a 50% relative increase. A chart should make clear which interpretation is intended.

8.2 Uncertainty displays

Estimates should be accompanied by uncertainty when uncertainty is decision-relevant. Error bars may represent standard deviations, standard errors, confidence intervals, prediction intervals, or credible intervals; these are not equivalent and must be labelled.

A confidence interval is a procedure with a long-run coverage interpretation under a specified model and sampling process. It is not strictly correct to say that a particular 95% confidence interval has a 95% probability of containing the fixed parameter. A Bayesian credible interval has a different interpretation based on a posterior distribution and prior assumptions. The visualization should identify the statistical meaning rather than using the generic phrase “error bars.”

8.3 Baselines and visual distortion

Truncating a bar chart can exaggerate differences. Conversely, forcing a line chart to begin at zero may compress meaningful variation when the line represents a rate or index and no length encoding is being used. The correct choice depends on the mark and the analytical task. The scale should be honest, labelled, and accompanied by direct values where interpretation could be ambiguous.

9. Avoiding Clutter and Misleading Conventions

Clutter arises when visual elements compete without adding analytical value. Common causes include excessive categories, redundant legends, unnecessary gridlines, long labels, multiple unrelated metrics, and decorative effects. The remedy is not always deletion; it may involve aggregation, faceting, filtering, annotation, or changing the chart type.

9.1 Small multiples and faceting

Small multiples repeat a consistent chart structure across groups. They support comparison by holding scales and encodings constant. Independent scales can reveal within-group patterns but weaken cross-group comparison; shared scales support comparison but may make small groups visually flat. The choice must be explicit.

9.2 Colour and accessibility

Use colour for a clear analytical purpose. Sequential palettes suit ordered magnitude, diverging palettes suit values around a meaningful midpoint, and qualitative palettes suit unordered categories. Avoid relying only on red–green distinctions because colour-vision deficiencies are common. Contrast, direct labelling, line styles, and texture can provide redundant cues.

9.3 Common misleading devices

Three-dimensional bars, tilted perspective, pictorial scaling, excessive smoothing, selective time windows, hidden missing values, and inconsistent denominators can all mislead. Starting a chart at an arbitrary date may conceal a longer trend. Showing only successful cases creates survivorship bias. A chart should disclose filters, exclusions, transformations, and data sources.

10. A Professional Workflow for Chart Design

A robust workflow can be organized as follows:

1. Define the decision. Identify who will use the display and what action or judgment it should support.

2. Formulate the analytical question. State whether the main task is comparison, distribution, relationship, composition, or ranking.

3. Audit the data. Check definitions, units, missingness, duplicates, outliers, denominators, and time coverage.

4. Select the visual form. Choose marks and channels that make the required comparison perceptually easy.

5. Establish scales and ordering. Use meaningful baselines, transformations, category order, and consistent units.

6. Add context. Include source, date, denominator, uncertainty, reference values, and essential annotations.

7. Reduce nonessential complexity. Remove decoration while preserving interpretation.

8. Validate. Compare the graphic with underlying data, test edge cases, inspect accessibility, and ask representative users to perform the intended task.

9. Document provenance. Record data version, transformations, code, assumptions, and publication date.

Validation should include both computational and perceptual checks. A chart can be numerically correct but visually ambiguous. Conversely, an attractive chart can contain a calculation error. Automated tests can verify totals, category mappings, and axis ranges; user testing can reveal whether readers infer the intended message.

11. Illustrative Case: Hospital Emergency Department Monitoring

Background

A regional hospital monitors emergency-department performance. Management wants to understand waiting times, compare service units, identify unusual days, and assess whether a new triage policy has changed performance.

Problem

The initial dashboard uses a pie chart for patient disposition, a dual-axis chart for arrivals and waiting time, and a bar chart of total complaints by unit. Reviewers discover that the busiest unit appears worst simply because it serves more patients, and the dual axes create an apparent relationship that changes when the scales are adjusted.

Analysis

The underlying data contain patient-level arrival and completion times, unit, acuity category, disposition, date, and patient volume. The relevant measures are median waiting time, upper-percentile waiting time, patient count, and complaint rate per 1,000 visits. The data also include repeated observations by day and strong weekday and seasonal effects.

Approach and solution

The redesign uses:

• a line chart of daily median waiting time with a clearly labelled rolling summary and a separate display of the 90th percentile;
• small-multiple dot plots comparing units, with confidence intervals or bootstrap intervals where appropriate;
• a 100-percent stacked bar chart for disposition shares, accompanied by a separate line or table for total patient volume;
• a sorted rate chart for complaints per 1,000 visits, showing denominators and minimum-volume warnings;
• a distribution plot of patient waiting times by acuity category;
• annotations marking the policy implementation date and periods of missing or changed data collection.

The dashboard uses a colour-blind-safe palette, direct labels, a common definition of waiting time, and filters that cannot silently change the denominator. A technical appendix records extraction queries and transformation rules.

Outcome

The redesign shows that the apparently worst unit has a high total number of complaints but a lower complaint rate than two smaller units. It also reveals that the policy reduced median waiting time for high-acuity patients but had little effect on the upper tail for low-acuity patients. The original dual-axis chart had obscured this distinction.

Learning points

The case demonstrates that chart selection must follow the decision question. Counts, rates, distributions, and trends answer different questions. Percentiles can reveal service risk that averages conceal; denominators are essential for fair comparisons; and uncertainty and data-quality annotations improve rather than weaken managerial interpretation.

12. Technology, AI, and Governance

Visualization software makes chart production easy, but automation does not guarantee analytical validity. Default chart recommendations may optimize for popularity or aesthetics rather than the user’s question. Automated systems can also select inappropriate aggregation, omit denominators, or infer a causal message from correlation.

AI-assisted visualization can help generate candidate charts, detect unusual values, write annotations, and translate code between tools. Its outputs should be treated as proposals requiring human review. Analysts must verify the data source, transformations, scale, statistical claims, and privacy implications. Sensitive dashboards should apply access controls, aggregation, suppression rules, and disclosure review.

Governance is especially important when charts inform employment, credit, healthcare, public safety, or resource allocation. A visualization can reproduce bias even when its arithmetic is correct if the underlying measurement system is unequal or incomplete. Documentation should identify who collected the data, which populations are missing, how categories were defined, and what decisions the chart is not suitable for.

13. Contemporary Developments and Best Practices

Interactive graphics allow filtering, zooming, drill-down, and linked views. Interaction can reduce initial clutter, but it can also hide the analytical frame. The default view should communicate the main message, and controls should preserve context by displaying active filters, units, denominators, and time ranges.

Uncertainty-aware visualization is increasingly important in forecasting, experimentation, climate analysis, and machine learning. Forecast fans, interval bands, calibration plots, residual displays, and sensitivity analyses help distinguish prediction from observation and estimate from fact. Fairness dashboards increasingly report performance across demographic groups, but subgroup comparisons require adequate sample sizes, consistent definitions, and attention to intersectional categories.

Best practice is therefore not a single catalogue of chart types. It is disciplined alignment among question, data, visual encoding, statistical assumptions, audience, and decision risk.

Chapter Summary

Chart selection begins with the analytical purpose and the structure of the data. Tables support precision and auditability; bar charts, dot plots, and slope charts support comparison and ranking; histograms, box plots, density plots, and ECDFs reveal distributions; scatter plots and time-series graphics display relationships; and stacked, pie, waterfall, treemap, and mosaic displays address composition under different conditions.

Position on a common scale is generally more precise than angle, area, or volume. Honest graphics require appropriate baselines, clearly labelled transformations, visible denominators, meaningful uncertainty indicators, and disclosure of missingness and data processing. Clutter can be reduced through ordering, small multiples, direct labelling, aggregation, and careful interaction. Ethical visualization requires attention to accessibility, privacy, bias, provenance, and the consequences of decisions based on the display.

A professional workflow defines the decision, formulates the question, audits the data, selects encodings, designs scales, adds context, validates both computation and perception, and documents provenance. AI and visualization software can accelerate production, but they do not replace statistical judgment or governance.

Key Terms

Analytical purpose: The decision or reasoning task a visualization is intended to support.

Bar chart: A chart using bars to compare quantities across discrete categories.

Baseline: The reference level from which a graphical mark is measured.

Composition: The structure of a whole in terms of its constituent parts.

Contingency table: A table showing counts or percentages for combinations of categorical variables.

Distribution: The pattern of values, including centre, spread, shape, and outliers.

ECDF: An empirical cumulative distribution function showing the proportion of observations at or below each value.

Encoding: The mapping of data values to visual properties.

Histogram: A display of quantitative values grouped into intervals.

IQR: Interquartile range, calculated as Q3 minus Q1.

Nominal variable: A categorical variable without an inherent order.

Overplotting: Visual overlap that conceals individual observations.

Ranking: Ordering items according to a measure.

Scatter plot: A display of paired quantitative observations using two position scales.

Sequential palette: A colour scheme representing ordered values from low to high.

Small multiples: Repeated, consistently designed charts for different groups or conditions.

Uncertainty interval: A graphical interval representing statistical uncertainty under a stated method.

Visual encoding: The use of position, length, colour, shape, or another channel to represent data.

Review Questions

Short Answer Questions

1. Why should chart selection begin with the analytical question rather than the available software template?

2. State two situations in which a table is preferable to a chart.

3. What information does a box plot communicate, and what may it conceal?

4. Why are denominators important when comparing counts across groups?

5. Distinguish between a percentage-point change and a relative percentage change.

Long Answer Questions

1. Compare bar charts, dot plots, and slope charts for ranking and comparison. Discuss their visual strengths and limitations.

2. Explain how bin width, smoothing, and sample size affect the interpretation of distribution graphics.

3. Discuss the statistical and perceptual risks of dual-axis charts, truncated bar charts, and three-dimensional graphics.

4. Develop a professional workflow for designing a dashboard that reports performance across organizational units.

5. Explain how uncertainty, accessibility, data provenance, and ethical governance should influence visualization design.

Analytical / Application Questions

1. A university compares student withdrawals across faculties using raw counts. Design a more appropriate visualization strategy, including denominators, uncertainty, and possible distributional displays.

2. A scatter plot of advertising expenditure and sales shows a strong upward association. Identify at least four additional analyses or design checks required before making a causal claim.

3. A management team wants to compare five product lines across revenue, margin, customer complaints, and growth. Propose a coordinated set of displays and justify the encoding, ordering, scales, and use of interaction.

Multiple Choice Questions with Answers

Question 1

Which visual encoding is generally most precise for quantitative comparison?

A. Position on a common scale

B. Area of a circle

C. Angle of a sector

D. Three-dimensional volume

Correct Answer: A. Position on a common scale

Question 2

A box plot commonly identifies potential outliers using which rule?

A. Values beyond one standard deviation from the mean

B. Values below Q1 − 1.5(IQR) or above Q3 + 1.5(IQR)

C. All values outside the median

D. Values above the maximum bin in a histogram

Correct Answer: B. Values below Q1 − 1.5(IQR) or above Q3 + 1.5(IQR)

Question 3

Which chart is most directly suited to displaying paired quantitative observations?

A. Pie chart

B. Scatter plot

C. Waterfall chart

D. Contingency table

Correct Answer: B. Scatter plot

Question 4

Why can raw complaint counts give an unfair comparison between service units?

A. Counts cannot be displayed in tables

B. Counts always contain sampling error

C. Units may serve different numbers of customers

D. Complaints are necessarily ordinal data

Correct Answer: C. Units may serve different numbers of customers

Question 5

What does IQR represent?

A. The mean divided by the range

B. The difference between Q3 and Q1

C. The distance between the minimum and maximum

D. The proportion above the median

Correct Answer: B. The difference between Q3 and Q1

Question 6

Which chart is usually appropriate for a time-ordered series?

A. Line chart

B. Pie chart

C. Treemap

D. Unsorted categorical bar chart

Correct Answer: A. Line chart

Question 7

What is a principal limitation of a 100-percent stacked bar chart?

A. It cannot show proportions

B. It removes information about absolute totals

C. It requires quantitative data only

D. It cannot compare categories

Correct Answer: B. It removes information about absolute totals

Question 8

Which colour scheme is generally appropriate for ordered values from low to high?

A. Sequential palette

B. Random qualitative palette

C. Binary palette with no legend

D. Unlabelled rainbow palette

Correct Answer: A. Sequential palette

Question 9

What is overplotting?

A. Using too few observations

B. Overlapping marks that conceal observations

C. Displaying a chart with a zero baseline

D. Sorting categories by value

Correct Answer: B. Overlapping marks that conceal observations

Question 10

What is a key governance requirement for an AI-generated chart?

A. Accepting the default chart without review

B. Removing the data source to simplify presentation

C. Verifying data, transformations, scales, and statistical claims

D. Using three-dimensional effects to show complexity

Correct Answer: C. Verifying data, transformations, scales, and statistical claims